Research On Recommendation Method Combining Tagging Context And Rating

Mining users' interests is the core and foundation of the recommendation.The general recommendation algorithm only reflects the user's interest preference by the user's rating on the project.With the development of Web 2.0,social tagging system allows users to add their own interests to the tag of any items,the tag as a keyword selected by the user,the user reflects the view of the project and interest,and then mark the dynamic changes of frequency and timing of the same time affects the label characterization of user interest.Therefore,combined with the Natural Science Fund Project “Research on social-driven context-aware personalized information service in ubiquitous computing environment”(Project No: 71471165),this paper study on the situation and rating with the recommended tagging method,systematically expounds the construction method combining recommendation model tagging and marking the situation,puts forward the index correlation based on frequency tagging and correlation strength index based on sequential mining user potential interest and study.The main contributions of this thesis are listed as follows:(1)Build the basic relationship model of users,tags and items.Through the user tagging behavior,we reduce the three-dimensional relationship among user-context-item into two two-dimensional relationships: user-context and context-item.We establish user item two-dimensional relationship through user's grading behavior,and intuitively show the relationship between the three in recommender system.(2)Put forward the tagging context correlation index.The correlation index is constructed with the Logistic population growth model,and the frequency of tagging is converted to correlation strength.According to the timing of user tagging behavior,the time forgetting curve is used to construct the correlation strength index.Obtaining users' interest preferences through these two-association metrics.(3)The recommendation method combining the tagging context and the rating.According to the timing behavior of user rating behavior,the rating is transformed by the correlation strength index,and then the recommendation model is built through the close relationship between the three users,tags and items.Get the user's estimate of the project in any tagging context,thus generating a list of recommendations.(4)Empirical research and analysis.This paper selects tagging data and rating data as research data set from representative social tagging application platform Movielens and evaluates and contrasts the three different evaluation indexes by using the accuracy evaluation rate,recall rate and F1 value.The research shows that the recommendation effect combining the tagging context and the rating is better than the recommendation effect that does not consider the tagging context.The frequency of the tagging cannot directly represent the close degree between the user and the context and the context and the item.Only the correlation index is the best recommendation effect.The rating behavior of the user is more stable,and it does not change with the change of time.The user's tagging behavior is influenced by time factor more than the time factor.The results of the study are of high practical application value for recommendation.